Verdicts

Every extraction carries a verdict. The model transcribes; deterministic code decides. Nothing about the verdict is the model’s opinion.

PASS The traverse closes and the computed area corroborates the area the document states. Both checks agree.

REVIEW Something needs human eyes, and the reasons say what. A misclosure past the threshold, an area that disagrees with the stated one, a course the extractor was unsure of, or a unit that could mean two things.

FAIL Not traversable. A call with no bearing, a unit the engine does not recognise, or a description that never located itself. The failing call is quoted.

The two checks

Closure, at 1:5,000. The traverse is recomputed from the extracted courses and walked back to its start. The gap between where it ends and where it began, over the perimeter, is the closure ratio.

Stated area, within 2%. The area the recomputed boundary encloses is compared against the acreage the document itself states, in the unit the document used.

They are independent on purpose. An extraction can close beautifully and enclose the wrong area, which is what a transposed digit looks like.

Why a REVIEW is worth as much as a PASS

The product is not the polygon, it is knowing which polygons to trust. A tool that quietly closes a failing traverse produces a drawing nobody downstream can question. A REVIEW that names its misclosure and quotes the suspect courses is the thing that was bought.

What the arithmetic cannot see

Closure and stated area can both be satisfied by a boundary that is mirrored, or that has a quadrant transposed, because those errors are symmetric. A reader can see it; the arithmetic cannot. That is why every certificate draws the tract.